The Linux graphics stack from X to Wayland (ars technica)
A key feature of Wayland is the use of a rendering API that does not have any dependencies on X. To maintain compatibility, the X Server itself is made into a Wayland client, and all of the X rendering is fed directly into Wayland. The Wayland package, like X before it, defines a protocol only. The architecture of Wayland, with its ability to function alongside X, provides an easy migration path for existing and even future planned X clients. The X Server can run as before, servicing all of the legacy clients."
Posted Mar 22, 2011 18:49 UTC (Tue)
by Thalience (subscriber, #4217)
[Link] (13 responses)
Posted Mar 22, 2011 19:46 UTC (Tue)
by rfunk (subscriber, #4054)
[Link] (2 responses)
In this case, X's own client/server terminology is a big part of the problem; the Ars writer is inconsistent about getting it right and getting it backward.
On the other hand, once I was able to sort some things out, it at least gave me a vague overview of what's happened in the handful of years since I last paid attention to the state of Linux graphics.
Posted Mar 22, 2011 20:24 UTC (Tue)
by drag (guest, #31333)
[Link]
Posted Mar 22, 2011 20:32 UTC (Tue)
by Thalience (subscriber, #4217)
[Link]
Posted Mar 22, 2011 20:35 UTC (Tue)
by PO8 (guest, #41661)
[Link] (2 responses)
Instead of this article, I would recommend Keith Packard's recent LCA talk (http://linuxconfau.blip.tv/file/4693305/). (Thanks to bersl2 on ars technica for the link.)
Posted Mar 23, 2011 13:40 UTC (Wed)
by roblucid (guest, #48964)
[Link] (1 responses)
Posted Mar 23, 2011 13:51 UTC (Wed)
by drag (guest, #31333)
[Link]
I was worried that I was making a big mistake due to the immaturity of open source drivers and I was dreading making the choice of frglx vs vesa drivers or something like that.
I am quite happy to report I am cranking away with a Radeon HD 5770 with Gallium drivers. It's not perfect, but it's surprisingly functional and stable.
I can't recommend it because this stuff is still very immature, but ATI open source drivers are doing fantastic.
Also don't be fooled by Phoronix benchmarks of Nexuiz and such. Many features are non-existent on OSS drivers, such as AA, so that some benchmarks are not comparing apples to apples.
Posted Mar 22, 2011 20:57 UTC (Tue)
by bk (guest, #25617)
[Link]
The article quality at Ars has gone down considerably since they were acquired by Conde Naste, unfortunately.
Posted Mar 22, 2011 23:18 UTC (Tue)
by renox (guest, #23785)
[Link] (1 responses)
Too bad, I used to really like Ars..
Posted Mar 23, 2011 9:34 UTC (Wed)
by k3ninho (subscriber, #50375)
[Link]
I think that this article needed to acknowledge that nVidia and AMD are competing, that nVidia's closed driver and CUDA libraries inspired AMD to respond by releasing the specs of their hardware. Also that AMD did that, hoping that the community would produce something which is native to Linux (I don't think I've not heard any announcements or code drops) and which is open and free.
K3n.
Posted Mar 23, 2011 9:31 UTC (Wed)
by wingo (guest, #26929)
[Link]
:)
Posted Mar 23, 2011 11:12 UTC (Wed)
by ebirdie (guest, #512)
[Link]
Posted Mar 23, 2011 14:37 UTC (Wed)
by sciurus (guest, #58832)
[Link]
Posted Mar 23, 2011 14:37 UTC (Wed)
by sciurus (guest, #58832)
[Link]
Posted Mar 22, 2011 21:40 UTC (Tue)
by yokem_55 (subscriber, #10498)
[Link] (4 responses)
Posted Mar 22, 2011 22:56 UTC (Tue)
by branden (guest, #7029)
[Link] (1 responses)
Posted Mar 22, 2011 23:06 UTC (Tue)
by jzbiciak (guest, #5246)
[Link]
Two Senators, even.
Posted Mar 23, 2011 2:36 UTC (Wed)
by aryonoco (guest, #55563)
[Link] (1 responses)
I asked him about Gwibber development on twitter, and he replied that he'll make an announcement about that soon, but it sounded like he's handing over its development to others.
In short, it looks like he has moved on from Open Source/Linux/Android, and if Evan Jenkins is going to be their new person for this position, he needs a lot of catching up to do to be able to replace Ryan Paul.
Posted Mar 23, 2011 12:40 UTC (Wed)
by kragil (guest, #34373)
[Link]
Exspect more articles of this quality or worse.
Posted Mar 23, 2011 10:00 UTC (Wed)
by dps (guest, #5725)
[Link] (6 responses)
Amoung other it is impossible to display most mac OS X applications on my 1920x1200 display, which is connected to a Linux box. If I was working from home it would be impossible to use those applications, period. X11 applications have none of these problems.
Posted Mar 23, 2011 11:37 UTC (Wed)
by ebirdie (guest, #512)
[Link] (2 responses)
+1 regarding remote graphics in OS X. However, what I have hoped to find in articles like this is, how Wayland and Spice are going to work together? The article a bit downplayed the remote X, yet again, like it was so yesterday obliviated by modern graphics hardware and the current mainstream of the whole graphics stack running locally is end of all there is and will be. I find it so, that graphics hardware have mostly little to do with, where I want to use a window ie. I'd like to keep LibreOffice window open at home and sometimes "draw" it to the desktop at workplace maintaining its session like a vim with screen in terminal. Of course X doesn't provide that per se either but when speaking about the functionality in the big picture of the graphics stack.
Posted Mar 23, 2011 18:18 UTC (Wed)
by drag (guest, #31333)
[Link] (1 responses)
X will have to be very well supported in Wayland as without it Wayland will be unusable. Nobody is going to port over applications in a hurry and X will remain to be the favored system for applications were portability is a high priority.
The reason a application developer would prefer to use Wayland instead of X is because they want to have good performance and need to make the full use of the video card for proper acceleration. For whatever reason. Wayland should make things like that much simpler. Which are the sort of things that would be so miserable over X that it would be unusable.
> However, what I have hoped to find in articles like this is, how Wayland and Spice are going to work together?
Well Spice works because of a paravirtualized driver combined with virtualization.
To make it work in a VirtualMachine you need to create a QXL driver for it. You can use regular generic VESA or VGA drivers and Spice works fine (far better then VNC or RDP) but for best performance you'd want to run a QXL driver.
If your dealing with non-virtualized systems then SPICE, as it is now, will not work.
Also there is no 'WAN' mode of SPICE yet. That's a 'TODO' and thus Apice is really designed for VDI deployments in a corporation or whatever.
For that, it's _FANTASTIC_.
For SPICE to work as a alternative to X for most things then you'd have to implement some sort of special Gallium Driver or kernel mode that implements the SPICE protocol or something like that.
Posted Mar 24, 2011 14:45 UTC (Thu)
by ebirdie (guest, #512)
[Link]
Good point.
> The reason a application developer would prefer to use Wayland instead of X is because they want to have good performance and need to make the full use of the video card for proper acceleration. For whatever reason.
How about the just released FF4's hookings into OpenGL as an example. What I can tell FF4 is using OpenGL hookings for drawing some specific graphics like Web3D and not the whole browser canvas. Whatever, what is my point and example with FF4 is, that I have found myself many times in situation, where I wanted to have my browser windows open at home desktop "drawn" onto my desktop at front, because I remembered that I had a particular tidbit open at home in a browser window. I have a bad habit of accumulating open browser windows and tabs and my associative memory works strangely, I remember things beside the tidpid I'm after, what makes searches a lot more tedious than getting quickly the view back. And many times the tidpid I'm after isn't googlable nor should be.
What you say sounds to me like a forking to avoid. If remote access or pulling of a graphic window were possible for a graphic drawing platform, I don't want to get into a situation that I have to remember "oh no, I had that checkmark for Wayland in FF preferences, can't pull the window, nothing can be done to get the tidbit except fly home". Not nice, especially if the "window pulling" were possible and mainstream in other "competing" platforms.
Posted Mar 23, 2011 17:03 UTC (Wed)
by nix (subscriber, #2304)
[Link] (2 responses)
Of course no such compositor exists yet and nobody capable of writing one appears to be interested in it...
Posted Mar 24, 2011 3:07 UTC (Thu)
by sciurus (guest, #58832)
[Link] (1 responses)
Posted Mar 24, 2011 4:22 UTC (Thu)
by amtota (guest, #4012)
[Link]
Posted Mar 23, 2011 11:57 UTC (Wed)
by xiobochang (guest, #73815)
[Link] (1 responses)
why whould anyone gnome with wayland
why whould red hat & novell break gnome
why does microsoft have 99% and linux 1%
gnome & red hat & novell are 1%
wot is deal with red hat & novell
we gnome dev now wot we talking
Posted Mar 23, 2011 23:59 UTC (Wed)
by jospoortvliet (guest, #33164)
[Link]
Or try to go to a FOSS conference & talk to someone there who knows what he/she is talking about...
Posted Mar 23, 2011 22:46 UTC (Wed)
by jcm (subscriber, #18262)
[Link] (18 responses)
Posted Mar 23, 2011 22:48 UTC (Wed)
by jcm (subscriber, #18262)
[Link]
Posted Mar 24, 2011 1:33 UTC (Thu)
by foom (subscriber, #14868)
[Link] (16 responses)
MS RDP always been way faster than X11, and these days you can do individual application sharing too, instead of just full-desktop. Maybe Wayland, since it's not bogged down by the historical baggage of the abysmally slow X11 remoting protocol will let us get to the point where Linux has a *good* remoting protocol instead.
And BTW, MS RDP also lets you trivially share printers, sound, and filesystems from your machine to the remote application...it sure would be nice if that was easily doable on linux too.
Here's a video I just found on YouTube demoing the basics of running a copy of MS Office remotely, in case you don't have windows handy:
Posted Mar 24, 2011 16:44 UTC (Thu)
by csigler (subscriber, #1224)
[Link] (13 responses)
I've used and have had good success with NX/NoMachine for remote work. From what I understand, MS did RDP "right." For Linux/Unix users, of course, RDP isn't a native solution. On those platforms, I'd suggest trying FreeNX.
Clemmitt
Posted Mar 24, 2011 17:15 UTC (Thu)
by jcm (subscriber, #18262)
[Link] (12 responses)
Jon.
Posted Mar 24, 2011 17:59 UTC (Thu)
by csigler (subscriber, #1224)
[Link] (3 responses)
NX/NoMachine is still rated as more-or-less the fastest solution (that I've found). Something equally tenable would be very nice. And yes, I like the native network-abstractability of X :) (My thanks to others for posting about xpra, etc.)
Clemmitt
Posted Mar 26, 2011 19:31 UTC (Sat)
by Wol (subscriber, #4433)
[Link] (2 responses)
And it's the HARDWARE that's the problem. TCP/IP is a servo-feedback protocol. Manufacturers have become so paranoid about losing packets (which tcp/ip is perfectly fine with), that instead they have massive, multiple-seconds-deep buffers. Which means via a heavily loaded router it can take many seconds for packets to get between machines. For something like X, which is interactive, it really kills the user experience if half your packets get there in microseconds, and others take tens of seconds!
If you're running telnet, that's PAINFUL.
If the hardware worked with, rather than against, TCP/IP, X would work fine and very zippily with maybe one tenth of current bandwidth! In other words X is fine with low bandwidth already.
Cheers,
Posted Mar 27, 2011 2:14 UTC (Sun)
by csigler (subscriber, #1224)
[Link] (1 responses)
OK, so the round-trip problem which NX addresses is in large part caused by the networking hardware 'twixt here and there. I appreciate this explanation; it's a good and succinct answer.
I guess the next logical question is, if X suffers from this round-trip latency problem -- even if it's not the X protocol's "fault" -- can (and will) a successor to X address this (to the point that the effect on performance is negligible)? Thanks again.
Clemmitt
Posted Mar 31, 2011 3:00 UTC (Thu)
by baldridgeec (guest, #55283)
[Link]
Some pro-Wayland people might say round trips are never necessary, to which I say that it's probably important to know if you couldn't allocate all the video ram you wanted to push this block. How do you find out? Error codes. How do you get them? Guess what.
On the other hand, I recently read the actual FAQ on the Wayland site and it reads much more lucidly than the rantings of fanboys (which is all the explanation I've been going on for months now.) But I still worry that it will end up blindly reimplementing X because the developers don't understand the purpose of its architecture.
FWIW, I use remote X all the time over the internet - even SSHing into one box so I can rdesktop into a Windows box at the datacenter. It works great for me, but of course that depends on your ISPs.
Posted Mar 24, 2011 18:00 UTC (Thu)
by njd27 (subscriber, #5770)
[Link] (5 responses)
If open source software happens because of having people scratch itches then I am confident that this one will get solved: because every time Wayland comes up there are so many people who seem to be so extremely itchy.
Posted Mar 24, 2011 18:29 UTC (Thu)
by nybble41 (subscriber, #55106)
[Link] (1 responses)
Of course, the intent is to retain compatibility for applications using the X protocol by running an X server on top of Wayland. However, this becomes less useful as the applications themselves are ported to interact with Wayland rather than X, and can no longer run on systems without local rendering capabilities.
I would be more sanguine about the changes if there was a way to make the underlying rendering protocol (e.g. Gallium3D) network-transparent, so that remote applications could make use of local rendering hardware. That doesn't seem like it would be very difficult, although it's hard to say what the performance would be like compared to the original remote X protocol (used correctly) or e.g. AIGLX.
Posted Mar 25, 2011 15:03 UTC (Fri)
by glisse (guest, #44837)
[Link]
As wayland protocol is also trying to design thing in way that an application can move from on wayland server to another you can then keep you wayland application running even if there is no wayland server (would be up to toolkit to stop rendering as there is no one to forward it too so it would just waste resource for nothing).
Really, wayland allow a much saner and better remotting than X ever did. Of course it moves the responsability to toolkit (for sane implementation). Others solution is to have a wayland server & compositor serving like a vnc server.
Posted Mar 25, 2011 4:11 UTC (Fri)
by JEDIDIAH (guest, #14504)
[Link] (1 responses)
This is a fundemental aspect of the system design. You can't just sweep it under the rug and either try to ignore it in some passive aggressive manner or just assume that it will be suitably dealt with "some how".
It's like trying to defer security considerations.
Others have pointed out what happens when you let stuff drop on the floor like that: the ugly mess that is using VNC for remotely accessing a Mac.
People need to stop taking design cues from Apple. Really. Do you people actually use their stuff? I have my doubts.
Posted Mar 26, 2011 7:46 UTC (Sat)
by Los__D (guest, #15263)
[Link]
Posted Mar 26, 2011 10:23 UTC (Sat)
by jcm (subscriber, #18262)
[Link]
Posted Mar 24, 2011 19:39 UTC (Thu)
by foom (subscriber, #14868)
[Link] (1 responses)
Posted Mar 26, 2011 6:47 UTC (Sat)
by ebirdie (guest, #512)
[Link]
Somewhere 2006 Adobe went even so far that they implemented a prevention into the Creative Suite installer of theirs preventing installation into Citrix Presentation Server (Presentation Server was the name for Metaframe later). Of course it was possible to circumvent the prevention, but for a company taking compliances with contracts and "best practices" seriously, this was no option.
Citrix's "remoting" technology was intresting and technically handy, but quite expensive, commercial intrests erupted it and finaly MS came and took it all offering "less expensive solution, since we have to have MS licenses and contracts anyway" in my superior's words.
Of course this tidbit of Citrix isn't anything for an itch to scratch at FOSS circles.
Posted Mar 24, 2011 20:40 UTC (Thu)
by pspinler (subscriber, #2922)
[Link] (1 responses)
Unfortunately, remote desktop still implies running a full GUI on the remote machine. In the case of all of my datacenter server hardware, I neither do this, nor desire to.
Note that this is not the same as saying that I never want to run a GUI config tool or application install, though. (Oracle, Websphere, DB2, you name it ... everything wants to do a graphic install these days. Bleah)
In fact, for all my virtual servers, I push the allocated memory down as much as possible, and would have to allocate more memory for each of several hundreds of vm's if I need to have a full remote GUI session running on each just in order to run a GUI tool on it.
Leave me remote GUI *application* execution, please.
-- Pat
Posted Mar 25, 2011 8:31 UTC (Fri)
by khim (subscriber, #9252)
[Link]
Sorry, but this is 10 years old trend - and it goes away. These days everything wants to do a web setup. You don't need either Wayland or X for this - just a local web browser and simple port forwarding.
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
I hope that they will be able to reverse the trend.
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
This sounds like how xpra works for X.The Linux graphics stack from X to Wayland (ars technica)
Xpra works by connecting to an ordinary X server as a compositing
manager -- but instead of combining the window images to present on
the screen, it takes the window images and stuffs them into a network
connection to the xpra client, which then displays them onto the
*remote* screen. It also acts as a window manager for the X server it
is running against, but it doesn't actually have any window manager
policy built into it. Instead, it takes all the window management
requests from the applications, sends them over the wire to the
client, who then issues those same requests on the real display, waits
to see what answer your real window manager gives, and then forwards
that answer back to the xpra server.
If you want recent packages of xpra and a GUI, try winswitch
(the only place where you'll find recent working xpra packages, including other platforms like win32 and OSX)
The Linux graphics stack from X to Wayland (ars technica)
With winswitch you can easily "send" windows from one machine to another.
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
http://www.youtube.com/watch?v=mmVbzEhGHJY
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
Wol
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland
The Linux graphics stack from X to Wayland
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
But MS RDP is not only a whole-desktop sharing solution. As the video showed, you can run an application (e.g. MS Word) on a remote machine, and have its UI appear naturally on the local desktop, just as if you were running it locally.
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
The Linux graphics stack from X to Wayland (ars technica)
Sorry, but this is solved problem...
Note that this is not the same as saying that I never want to run a GUI config tool or application install, though. (Oracle, Websphere, DB2, you name it ... everything wants to do a graphic install these days. Bleah)